Low-dose-rate brachytherapy and long-term treatment outcomes in patients younger than 60 years of age.

Purpose: Low-dose-rate (LDR) brachytherapy in young men remains controversial amongst urologists due to their concerns regarding long-term biochemical control and treatment-related toxicities. The purpose of this study was to evaluate the treatment outcomes of men under 60 years of age who underwent LDR brachytherapy with iodine-125 (125I) for clinically localized low- to intermediate-risk prostate cancer. Material and methods: All consecutive patients with clinically localized prostate cancer treated at our institution from 2003 to 2016 with 125I monotherapy were included in the study. Prescription dose was 145.0 Gy modified peripheral loading (MPD). All patients were assessed for biochemical progression-free survival using Phoenix definition (nadir +2 ng/ml), clinical progression-free survival, overall survival (OS), and any associated treatment toxicity. Results: A total of 161 patients were included, with a median follow-up of 6.8 years (range, 3-14.54 years). Median age at implant was 57 years (range, 53-59 years). Mean prostate specific antigen (PSA) level at diagnosis was 4.43 ng/ml (SD = 2.29). Majority of men had low-risk prostate cancer (70.2%). Biochemical progression-free survival at 8 years was 94% for the entire cohort. Median PSA at 4 years was 0.169 (IQR, 0.096-0.360), with 45% of patients having a PSA greater than 0.2. OS was 96.9%, with 5 deaths reported but only one was secondary to prostate cancer. Late grade > 2 genitourinary toxicities were reported in 18 patients (11.2%). Three patients (1.9%) developed secondary cancers, all considered unrelated to their LDR brachytherapy. Conclusions: With excellent long-term treatment outcomes and minimal associated toxicities, our results showed that LDR brachytherapy can be an effective treatment of choice in younger men.

Analytical and Functional Similarity of Aflibercept Biosimilar ABP 938 with Aflibercept Reference Product.

INTRODUCTION: ABP 938 is being developed as a biosimilar candidate to aflibercept reference product (RP), a biologic used for certain angiogenic eye disorders. This study was designed to provide a comparative analytical assessment of the structural and functional attributes of ABP 938 and aflibercept RP sourced from the United States (US) and the European Union (EU). METHODS: Structural and functional characterization studies were performed using state-of-the-art analytical techniques that were appropriate to assess relevant quality attributes and capable of detecting qualitative and quantitative differences in primary structure, higher-order structure and biophysical properties, product-related substances and impurities, general properties, and biological activities. RESULTS: ABP 938 had the same amino acid sequence and exhibited similar secondary and tertiary structures, and biological activity as aflibercept RP. There were minor differences in a small number of biochemical attributes which are not expected to impact clinical performance. In addition, aflibercept RP sourced from the US and EU were analytically similar. CONCLUSIONS: ABP 938 was structurally and functionally similar to aflibercept RP. Since aflibercept RP sourced from the US and EU were analytically similar, this allows for the development of a scientific bridge such that a single-source RP can be used in nonclinical and clinical studies.

Eylea^® (aflibercept) is a biologic medication approved for the treatment of patients with certain eye diseases that can result in low vision or blindness. Biosimilars are biologic medications that are highly similar to an existing approved biologic medication, often called a reference product. Biosimilars have the potential to reduce medication costs despite having no clinically significant differences in quality, efficacy, and safety from their reference products. ABP 938 is currently being developed as a biosimilar to aflibercept reference product. We have conducted similarity studies to compare multiple batches of ABP 938 and aflibercept reference product sourced from both the United States and the European Union, using state-of-the-art analytical methods. The results demonstrated that ABP 938 had the same amino acid sequence and similar structural and biological activities as aflibercept reference product. Before biosimilars can be used as medicines, studies such as this one are required by the Food and Drug Administration and other regulatory authorities to ensure that biosimilars are as safe and effective as their reference products.

A single cell atlas of frozen shoulder capsule identifies features associated with inflammatory fibrosis resolution.

Frozen shoulder is a spontaneously self-resolving chronic inflammatory fibrotic human disease, which distinguishes the condition from most fibrotic diseases that are progressive and irreversible. Using single-cell analysis, we identify pro-inflammatory MERTKlowCD48+ macrophages and MERTK + LYVE1 + MRC1+ macrophages enriched for negative regulators of inflammation which co-exist in frozen shoulder capsule tissues. Micro-cultures of patient-derived cells identify integrin-mediated cell-matrix interactions between MERTK+ macrophages and pro-resolving DKK3+ and POSTN+ fibroblasts, suggesting that matrix remodelling plays a role in frozen shoulder resolution. Cross-tissue analysis reveals a shared gene expression cassette between shoulder capsule MERTK+ macrophages and a respective population enriched in synovial tissues of rheumatoid arthritis patients in disease remission, supporting the concept that MERTK+ macrophages mediate resolution of inflammation and fibrosis. Single-cell transcriptomic profiling and spatial analysis of human foetal shoulder tissues identify MERTK + LYVE1 + MRC1+ macrophages and DKK3+ and POSTN+ fibroblast populations analogous to those in frozen shoulder, suggesting that the template to resolve fibrosis is established during shoulder development. Crosstalk between MerTK+ macrophages and pro-resolving DKK3+ and POSTN+ fibroblasts could facilitate resolution of frozen shoulder, providing a basis for potential therapeutic resolution of persistent fibrotic diseases.

Fluorine-18-labelled Prostate-specific Membrane Antigen Positron Emission Tomography/Computed Tomography or Magnetic Resonance Imaging to Diagnose and Localise Prostate Cancer. A Prospective Single-arm Paired Comparison (PEDAL).

BACKGROUND AND OBJECTIVE: Multiparametric magnetic resonance imaging (mpMRI) of the prostate is used for prostate cancer diagnosis. However, mpMRI has lower sensitivity for small tumours. Prostate-specific membrane antigen positron emission tomography/computed tomography (PSMA-PET/CT) offers increased sensitivity over conventional imaging. This study aims to determine whether the diagnostic accuracy of 18F-DCFPyL PSMA-PET/CT was superior to that of mpMRI for detecting prostate cancer (PCa) at biopsy. METHODS: Between 2020 and 2021, a prospective multicentre single-arm phase 3 imaging trial enrolled patients with clinical suspicion for PCa to have both mpMRI and PSMA-PET/CT (thorax to thigh), with reviewers blinded to the results of other imaging. Multiparametric MRI was considered positive for Prostate Imaging Reporting and Data System (PIRADS) 3-5. PSMA-PET/CT was assessed quantitatively (positive maximum standardised uptake value [SUVmax] >7) and qualitatively (five-point lexicon of certainty). Patients underwent targeted and systematic biopsy, with the technique at the discretion of the treating urologist. Clinically significant PCa (csPCa) was defined as International Society of Urological Pathology grade group (GG) ≥2. The primary outcome was the diagnostic accuracy for detecting PCa, reported as sensitivity, specificity, negative predictive value (NPV), and area under the curve (AUC) of the receiver operating curve. The secondary endpoints included a comparison of the diagnostic accuracy for detecting csPCa, assessing gains in combining PMSA-PET/CT with mpMRI to mpMRI alone. KEY FINDINGS AND LIMITATIONS: Of the 236 patients completing both mpMRI and PSMA-PET/CT, 184 (76.7%) had biopsy. Biopsy histology was benign (n = 73), GG 1 (n = 27), and GG ≥2 (n = 84). The diagnostic accuracy of mpMRI for detecting PCa (AUC 0.76; 95% confidence interval [CI] 0.69, 0.82) was higher than that of PSMA-PET/CT (AUC 0.63; 95% CI 0.56, 0.70, p = 0.03). The diagnostic accuracy of mpMRI for detecting csPCa (AUC 0.72; 95% CI 0.67, 0.78) was higher than that of PSMA-PET/CT (AUC 0.62; 95% CI 0.55, 0.69) but not statistically significant (p = 0.27). A combination of PSMA-PET/CT and mpMRI showed excellent sensitivity (98.8%, 95% CI 93.5%, 100%) and NPV (96%, 95% CI 79.6%, 99.9%) over mpMRI alone (86.9% and 80.7%, respectively, p = 0.01). Thirty-two patients (13.6%) had metastatic disease. They tended to be older (68.4 vs 65.1 yr, p = 0.023), and have higher prostate-specific antigen (PSA; median PSA 9.6 vs 6.2ng/ml, p < 0.001) and abnormal prostate on digital rectal examination (78.2% vs 44.1%, p < 0.001). CONCLUSIONS AND CLINICAL IMPLICATIONS: Multiparametric MRI had superior diagnostic accuracy to PSMA-PET/CT for detecting PCa, though the difference is not significant in case of csPCa detection. A combination of mpMRI and PSMA-PET/CT showed improved sensitivity and NPV. PSMA-PET/CT could be considered for diagnostic use in patients unable to have mpMRI or those with concerning clinical features but negative mpMRI. PATIENT SUMMARY: In this trial, we compared the ability of 18F-labelled prostate-specific membrane antigen positron emission tomography/computed tomography (PSMA-PET/CT) with that of multiparametric magnetic resonance imaging (mpMRI) to diagnose prostate cancer by biopsy in a prostate-specific antigen screening population. We found that MRI was superior to PSMA to diagnose prostate cancer, though there was no difference in ability to diagnose clinically significant prostate cancer. PSMA-PET/CT could be considered for diagnostic use in patients unable to have mpMRI or those with concerning clinical features but negative mpMRI. Combining MRI with PSMA-PET increases the negative predictive value over MRI alone and may help men avoid invasive prostate biopsy.

A genome-wide association meta-analysis implicates Hedgehog and Notch signaling in Dupuytren's disease.

Dupuytren's disease (DD) is a highly heritable fibrotic disorder of the hand with incompletely understood etiology. A number of genetic loci, including Wnt signaling members, have been previously identified. Our overall aim was to identify novel genetic loci, to prioritize genes within the loci for functional studies, and to assess genetic correlation with associated disorders. We performed a meta-analysis of six DD genome-wide association studies from three European countries and extensive bioinformatic follow-up analyses. Leveraging 11,320 cases and 47,023 controls, we identified 85 genome-wide significant single nucleotide polymorphisms in 56 loci, of which 11 were novel, explaining 13.3-38.1% of disease variance. Gene prioritization implicated the Hedgehog and Notch signaling pathways. We also identified a significant genetic correlation with frozen shoulder. The pathways identified highlight the potential for new therapeutic targets and provide a basis for additional mechanistic studies for a common disorder that can severely impact hand function.

Low-Rank Tensor Function Representation for Multi-Dimensional Data Recovery.

Since higher-order tensors are naturally suitable for representing multi-dimensional data in real-world, e.g., color images and videos, low-rank tensor representation has become one of the emerging areas in machine learning and computer vision. However, classical low-rank tensor representations can solely represent multi-dimensional discrete data on meshgrid, which hinders their potential applicability in many scenarios beyond meshgrid. To break this barrier, we propose a low-rank tensor function representation (LRTFR) parameterized by multilayer perceptrons (MLPs), which can continuously represent data beyond meshgrid with powerful representation abilities. Specifically, the suggested tensor function, which maps an arbitrary coordinate to the corresponding value, can continuously represent data in an infinite real space. Parallel to discrete tensors, we develop two fundamental concepts for tensor functions, i.e., the tensor function rank and low-rank tensor function factorization, and utilize MLPs to paramterize factor functions of the tensor function factorization. We theoretically justify that both low-rank and smooth regularizations are harmoniously unified in LRTFR, which leads to high effectiveness and efficiency for data continuous representation. Extensive multi-dimensional data recovery applications arising from image processing (image inpainting and denoising), machine learning (hyperparameter optimization), and computer graphics (point cloud upsampling) substantiate the superiority and versatility of our method as compared with state-of-the-art methods. Especially, the experiments beyond the original meshgrid resolution (hyperparameter optimization) or even beyond meshgrid (point cloud upsampling) validate the favorable performances of our method for continuous representation.

Polygenic Risk Associations with Clinical Characteristics and Recurrence of Dupuytren Disease.

BACKGROUND: Dupuytren disease (DD) is a common complex trait, with varying severity and incompletely understood cause. Genome-wide association studies (GWAS) have identified risk loci. In this article, we examine whether genetic risk profiles of DD in patients are associated with clinical variation and disease severity and with patient genetic risk profiles of genetically correlated traits, including body mass index (BMI), triglycerides, high-density lipoproteins, type 2 diabetes mellitus, and endophenotypes fasting glucose and glycated hemoglobin. METHODS: The authors used a well-characterized cohort of 1461 DD patients with available phenotypic and genetic data. Phenotype data include age at onset, recurrence, and family history of disease. Polygenic risk scores (PRSs) of DD, BMI, triglycerides, high-density lipoprotein, type 2 diabetes, fasting glucose, and hemoglobin A1c using various significance thresholds were calculated with PRSice using the most recent GWAS summary statistics. Control data from LifeLines were used to determine P value cutoffs for PRS generation explaining most variance. RESULTS: The PRS for DD was significantly associated with a positive family history for DD, age at onset, disease onset before the age of 50, and recurrence. We also found a significant negative correlation between the PRSs for DD and BMI. CONCLUSIONS: Although GWAS studies of DD are designed to identify genetic risk factors distinguishing case/control status, we show that the genetic risk profile for DD also explains part of its clinical variation and disease severity. The PRS may therefore aid in accurate prognostication, choosing initial treatment and in personalized medicine in the future. CLINICAL QUESTION/LEVEL OF EVIDENCE: Risk, III.

Transfer Learning With Singular Value Decomposition of Multichannel Convolution Matrices.

The task of transfer learning using pretrained convolutional neural networks is considered. We propose a convolution-SVD layer to analyze the convolution operators with a singular value decomposition computed in the Fourier domain. Singular vectors extracted from the source domain are transferred to the target domain, whereas the singular values are fine-tuned with a target data set. In this way, dimension reduction is achieved to avoid overfitting, while some flexibility to fine-tune the convolution kernels is maintained. We extend an existing convolution kernel reconstruction algorithm to allow for a reconstruction from an arbitrary set of learned singular values. A generalization bound for a single convolution-SVD layer is devised to show the consistency between training and testing errors. We further introduce a notion of transfer learning gap. We prove that the testing error for a single convolution-SVD layer is bounded in terms of the gap, which motivates us to develop a regularization model with the gap as the regularizer. Numerical experiments are conducted to demonstrate the superiority of the proposed model in solving classification problems and the influence of various parameters. In particular, the regularization is shown to yield a significantly higher prediction accuracy.

Eigenimage2Eigenimage (E2E): A Self-Supervised Deep Learning Network for Hyperspectral Image Denoising.

The performance of deep learning-based denoisers highly depends on the quantity and quality of training data. However, paired noisy-clean training images are generally unavailable in hyperspectral remote sensing areas. To solve this problem, this work resorts to the self-supervised learning technique, where our proposed model can train itself to learn one part of noisy input from another part of noisy input. We study a general hyperspectral image (HSI) denoising framework, called Eigenimage2Eigenimage (E2E), which turns the HSI denoising problem into an eigenimage (i.e., the subspace representation coefficients of the HSI) denoising problem and proposes a learning strategy to generate noisy-noisy paired training eigenimages from noisy eigenimages. Consequently, the E2E denoising framework can be trained without clean data and applied to denoise HSIs without the constraint with the number of frequency bands. Experimental results are provided to demonstrate the performance of the proposed method that is better than the other existing deep learning methods for denoising HSIs. A MATLAB demo of this work is available at https://github.com/LinaZhuang/HSI-denoiser-Eigenimage2Eigenimagehttps://github.com/LinaZhuang/HSI-denoiser-Eigenimage2Eigenimage for the sake of reproducibility.

Re-defining 18 F-DCFPyl prostate-specific membrane antigen positron emission tomography detected local recurrence from radical prostatectomy histopathology.

OBJECTIVE: To assess topographic concordance between the histopathological features of patients' radical prostatectomy (RP) specimens and the location of the prostate-specific membrane antigen positron emission tomography (PSMA PET) local recurrences, qualitatively and quantitatively. PATIENTS AND METHODS: Our cohort was selected from the 100 men who received a 18 F-DCFPyL PET scan in the IMPPORT trial (Australian New Zealand Clinical Trials Registry Number: ACTRN12618001530213), a prospective non-randomised study completed by GenesisCare Victoria. Eligibility included patients with a rising prostate-specific antigen (PSA) level (>0.2 ng/mL) after RP and PSMA PET detected local recurrence. Histopathological parameters collated included the location of tumour, extraprostatic extension (EPE), and positive margins. Criteria for the location and 'concordance' between histopathological features and local recurrences were pre-defined. RESULTS: A total of 24 patients were eligible; the median age was 71 years, the median PSA level was 0.37 ng/mL, and the time between RP and PSMA PET was 2.6 years. In all, 15 patients had recurrences within the vesicourethral anastomotic region and nine within the lateral surgical margins. There was 100% concordance in the left-right plane between tumour location and local recurrence, with 79% of these lesions concordant three-dimensionally; across craniocaudal, left-right, and anterior-posterior planes. In all, 10 of the 16 (63%) patients with EPE and five of the nine patients with positive margins had three-dimensional concordance between their pathology and their local recurrence. In quantitative assessment, 17 of the 24 patients, had local recurrences that correlated with the location of their original tumour in the craniocaudal plane. CONCLUSION: Local recurrence is highly concordant with the position of the tumour within the prostate. Predicting the location of local recurrence using the location of the EPE and positive margins is less helpful. Further investigation into this field, could impact surgical technique and salvage radiotherapy clinical target volume.

Multi-relational graph convolutional networks: Generalization guarantees and experiments.

The class of multi-relational graph convolutional networks (MRGCNs) is a recent extension of standard graph convolutional networks (GCNs) to handle heterogenous graphs with multiple types of relationships. MRGCNs have been shown to yield results superior than traditional GCNs in various machine learning tasks. The key idea is to introduce a new kind of convolution operated on tensors that can effectively exploit correlations exhibited in multiple relationships. The main objective of this paper is to analyze the algorithmic stability and generalization guarantees of MRGCNs to confirm the usefulness of MRGCNs. Our contributions are of three folds. First, we develop a matrix representation of various tensor operations underneath MRGCNs to simplify the analysis significantly. Next, we prove the uniform stability of MRGCNs and deduce the convergence of the generalization gap to support the usefulness of MRGCNs. The analysis sheds lights on the design of MRGCNs, for instance, how the data should be scaled to achieve the uniform stability of the learning process. Finally, we provide experimental results to demonstrate the stability results.

Tucker network: Expressive power and comparison.

Deep neural networks have achieved great success in solving many machine learning and computer vision problems. In this paper, we propose a deep neural network called the Tucker network derived from the Tucker format and analyze its expressive power. The results demonstrate that the Tucker network has exponentially higher expressive power than the shallow network. In other words, a shallow network with an exponential width is required to realize the same score function as that computed by the Tucker network. Moreover, we discuss the expressive power between the hierarchical Tucker tensor network (HT network) and the proposed Tucker network. To generalize the Tucker network into a deep version, we combine the hierarchical Tucker format and Tucker format to propose a deep Tucker tensor decomposition. Its corresponding deep Tucker network is presented. Experiments are conducted on three datasets: MNIST, CIFAR-10 and CIFAR-100. The results experimentally validate the theoretical results and show that the Tucker network and deep Tucker network have better performance than the shallow network and HT network.

PSMA-PET Detection of Unusual Metastases in Castrate-Sensitive Prostate Carcinoma.

ABSTRACT: Prostate cancer (PCa) is a multifaceted, heterogeneous disease (with 7 molecular subtypes), which can metastasize to common sites, such as bone, lymph nodes, liver, and lungs. However, with PSMA PET imaging, rare sites of metastasis are increasingly discovered. We report 5 cases of unusual metastases in patients with castrate-sensitive PCa: solitary right inguinal nodal metastasis, solitary abdominal wall metastasis, penile shaft metastases, solitary perineum metastasis, and pleural metastases. These cases further support the use of PSMA-PET imaging in PCa monitoring, with the ability to detect solitary, small volume, and rare sites of metastases, which may not be apparent on conventional imaging.

Development of Molecular Magnetic Resonance Imaging Tools for Longitudinal Tracking of Carotid Atherosclerotic Disease Using Fast Imaging with Steady-State Precession.

Identification of patients with high-risk asymptomatic carotid plaques remains a challenging but essential step in stroke prevention. Current selection criteria for intervention in carotid disease are still determined by symptomatology and degree of luminal stenosis. This strategy has been less effective in identifying the high-risk asymptomatic individual patients. Inflammation is the key factor that drives plaque instability causing clinical sequelae. Currently, there is no imaging tool in routine clinical practice to assess the inflammatory status within atherosclerotic plaques. Herein we describe the development of a novel molecular magnetic resonance imaging (MRI) strategy to interrogate plaque inflammation, and hence its vulnerability in vivo, using dual-targeted iron particle-based probes and fast imaging with steady-state precession (FISP) sequence, adding further prognostic information to luminal stenosis alone. A periarterial cuff was used to generate high-risk plaques at specific timepoints and location of the carotid artery in an apolipoprotein-E-deficient mouse model. Using this platform, we demonstrated that in vivo dual-targeted iron particles with enhanced FISP can (i) target and characterise high-risk vulnerable plaques and (ii) quantitatively report and track the inflammatory activity within carotid plaques longitudinally. This molecular imaging tool may permit (i) accurate monitoring of the risk of carotid plaques and (ii) timely identification of high-risk asymptomatic patients for prophylactic carotid intervention, achieving early stroke prevention.

Effect of physical exercise on fear of falling in patients with stroke: A systematic review and meta-analysis.

OBJECTIVE: To consolidate the evidence on the effect of physical exercise on fear of falling in individuals with stroke. DATA SOURCES: PubMed, CINAHL, Cochrane Database and MEDLINE. METHODS: An extensive database search was conducted to identify the randomised controlled trials that examined the effect of physical exercise on fear of falling post-stroke. Grading of Recommendation, Assessment, Development and Evaluation (GRADE) was used to assess the quality of evidence for each meta-analysis. RESULTS: Fourteen trials totalling 1211 participants were included in this review. Thirteen of these (1180 participants) were included in the meta-analyses. In the primary analysis, very low-quality evidence suggested that exercise reduced fear of falling post-stroke (standardized mean difference (SMD) 0.48; 95% confidence interval (CI) 0.23 to 0.72). The effect was diminished at three- to six-month follow-up after exercise training ended (SMD -0.09; 95% CI -0.27 to 0.10; high-quality evidence). In the sensitivity analyses, the treatment effect was more pronounced in individuals with a lower baseline Berg balance score (BBS ≤45; SMD 0.53; 95%CI 0.17 to 0.88) and for those trials with exercise frequency of ≥3 sessions per week (SMD 0.70; 95%CI 0.39 to 1.01). Compared with circuit-based training consisting of a combination of walking, balance and strengthening exercises (SMD 0.27; 95% CI -0.09 to 0.63), walking programmes seemed to generate a larger effect on fear of falling (SMD 1.06; 95%CI 0.43 to 1.70). CONCLUSION: Physical exercise was beneficial for reducing fear of falling in individuals with stroke, particularly those with poorer balance ability.

Dictionary Learning With Low-Rank Coding Coefficients for Tensor Completion.

In this article, we propose a novel tensor learning and coding model for third-order data completion. The aim of our model is to learn a data-adaptive dictionary from given observations and determine the coding coefficients of third-order tensor tubes. In the completion process, we minimize the low-rankness of each tensor slice containing the coding coefficients. By comparison with the traditional predefined transform basis, the advantages of the proposed model are that: 1) the dictionary can be learned based on the given data observations so that the basis can be more adaptively and accurately constructed and 2) the low-rankness of the coding coefficients can allow the linear combination of dictionary features more effectively. Also we develop a multiblock proximal alternating minimization algorithm for solving such tensor learning and coding model and show that the sequence generated by the algorithm can globally converge to a critical point. Extensive experimental results for real datasets such as videos, hyperspectral images, and traffic data are reported to demonstrate these advantages and show that the performance of the proposed tensor learning and coding method is significantly better than the other tensor completion methods in terms of several evaluation metrics.

FastHyMix: Fast and Parameter-Free Hyperspectral Image Mixed Noise Removal.

The decrease in the widths of spectral bands in hyperspectral imaging leads to a decrease in signal-to-noise ratio (SNR) of measurements. The decreased SNR reduces the reliability of measured features or information extracted from hyperspectral images (HSIs). Furthermore, the image degradations linked with various mechanisms also result in different types of noise, such as Gaussian noise, impulse noise, deadlines, and stripes. This article introduces a fast and parameter-free hyperspectral image mixed noise removal method (termed FastHyMix), which characterizes the complex distribution of mixed noise by using a Gaussian mixture model and exploits two main characteristics of hyperspectral data, namely, low rankness in the spectral domain and high correlation in the spatial domain. The Gaussian mixture model enables us to make a good estimation of Gaussian noise intensity and the locations of sparse noise. The proposed method takes advantage of the low rankness using subspace representation and the spatial correlation of HSIs by adding a powerful deep image prior, which is extracted from a neural denoising network. An exhaustive array of experiments and comparisons with state-of-the-art denoisers was carried out. The experimental results show significant improvement in both synthetic and real datasets. A MATLAB demo of this work is available at https://github.com/LinaZhuang for the sake of reproducibility.

Morphological Feature Visualization of Alzheimer's Disease via Multidirectional Perception GAN.

The diagnosis of early stages of Alzheimer's disease (AD) is essential for timely treatment to slow further deterioration. Visualizing the morphological features for early stages of AD is of great clinical value. In this work, a novel multidirectional perception generative adversarial network (MP-GAN) is proposed to visualize the morphological features indicating the severity of AD for patients of different stages. Specifically, by introducing a novel multidirectional mapping mechanism into the model, the proposed MP-GAN can capture the salient global features efficiently. Thus, using the class discriminative map from the generator, the proposed model can clearly delineate the subtle lesions via MR image transformations between the source domain and the predefined target domain. Besides, by integrating the adversarial loss, classification loss, cycle consistency loss, and L1 penalty, a single generator in MP-GAN can learn the class discriminative maps for multiple classes. Extensive experimental results on Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset demonstrate that MP-GAN achieves superior performance compared with the existing methods. The lesions visualized by MP-GAN are also consistent with what clinicians observe.

Hypergraph Collaborative Network on Vertices and Hyperedges.

In many practical datasets, such as co-citation and co-authorship, relationships across the samples are more complex than pair-wise. Hypergraphs provide a flexible and natural representation for such complex correlations and thus obtain increasing attention in the machine learning and data mining communities. Existing deep learning-based hypergraph approaches seek to learn the latent vertex representations based on either vertices or hyperedges from previous layers and focus on reducing the cross-entropy error over labeled vertices to obtain a classifier. In this paper, we propose a novel model called Hypergraph Collaborative Network (HCoN), which takes the information from both previous vertices and hyperedges into consideration to achieve informative latent representations and further introduces the hypergraph reconstruction error as a regularizer to learn an effective classifier. We evaluate the proposed method on two cases, i.e., semi-supervised vertex and hyperedge classifications. We carry out the experiments on several benchmark datasets and compare our method with several state-of-the-art approaches. Experimental results demonstrate that the performance of the proposed method is better than that of the baseline methods.

Shared genetic architecture of hernias: A genome-wide association study with multivariable meta-analysis of multiple hernia phenotypes.

Abdominal hernias are common and characterised by the abnormal protrusion of a viscus through the wall of the abdominal cavity. The global incidence is 18.5 million annually and there are limited non-surgical treatments. To improve understanding of common hernia aetiopathology, we performed a six-stage genome-wide association study (GWAS) of 62,637 UK Biobank participants with either single or multiple hernia phenotypes including inguinal, femoral, umbilical and hiatus hernia. Additionally, we performed multivariable meta-analysis with metaUSAT, to allow integration of summary data across traits to generate combined effect estimates. On individual hernia analysis, we identified 3404 variants across 38 genome-wide significant (p < 5×10-8) loci of which 11 are previously unreported. Robust evidence for five shared susceptibility loci was discovered: ZC3H11B, EFEMP1, MHC region, WT1 and CALD1. Combined hernia phenotype analyses with additional multivariable meta-analysis of summary statistics in metaUSAT revealed 28 independent (seven previously unreported) shared susceptibility loci. These clustered in functional categories related to connective tissue and elastic fibre homeostasis. Weighted genetic risk scores also correlated with disease severity suggesting a phenotypic-genotypic severity correlation, an important finding to inform future personalised therapeutic approaches to hernia.